Model Ruler — AI Cost Calculators

Measured conclusion: Deep-tested. L2 outcome: tools_ok.

Read-only AI/LLM cost tools with current MCP discovery and attributable Registry transport.

Add to your agent

Copy this into your MCP client config:

{
  "mcpServers": {
    "Model Ruler — AI Cost Calculators": {
      "url": "https://modelruler.dev/mcp?dist=model_everywhere_registry_native_official_registry",
      "transport": "streamable-http"
    }
  }
}

Measured tools (12)

• token-counter read-only

  Use when a user asks how many tokens a given text will consume, or needs to estimate prompt size before pricing a workload. Given text and tokenizer family, returns low/high token range and byte-level measurements.

  Parameters:

  • text (required) — string — Text content to count tokens for
  • tokenizer — string — Tokenizer family (default: default)
  • expected_out_tokens — number — Expected output token budget (optional)

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• provider-cost-calculator read-only

  Use when a user asks what an LLM workload costs on a specific provider/model, or wants to compare cost across providers. Given tokens per call and call volume, returns monthly cost plus a tier comparison table.

  Parameters:

  • tokens_in (required) — number — Input tokens per call
  • tokens_out (required) — number — Output tokens per call
  • model — string — Target model (e.g. claude-sonnet-4-6)
  • provider — string — Target provider (e.g. anthropic, openai, together)
  • calls_per_month — number — Monthly call volume
  • include_comparison — boolean — Include tier comparison table (default true)

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• self-host-breakeven-calculator read-only

  Use when a user is deciding between API usage and self-hosted GPU inference at a given volume. Returns breakeven token volume, monthly cost comparison, and go/no-go recommendation.

  Parameters:

  • monthly_tokens (required) — number — Monthly output token volume
  • gpu_type — string — GPU type (e.g. h100, a100-80gb)
  • gpu_provider — string — GPU provider (e.g. runpod, modal)
  • utilization_pct — number — Expected GPU utilization % (default 60)
  • api_cost_per_1m_out — number — Current API output cost per 1M tokens
  • operational_overhead_pct — number — Ops overhead % on GPU cost (default 40)

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• context-window-planner read-only

  Use when a user needs to know whether a document plus prompt plus output fits within a model's context window, or wants a strategy recommendation (truncate/summarize/rag/chunk).

  Parameters:

  • doc_tokens (required) — number — Primary document/content tokens
  • model_context_window (required) — number — Target model context size
  • strategy_hint — string — Preferred strategy (optional)
  • overhead_tokens — number — System prompt + few-shot + history (default 2000)
  • expected_out_tokens — number — Reserved output budget (default 1000)

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• quantization-calculator read-only

  Use when a user is planning to quantize an LLM to fit on smaller hardware. Given parameter count and precision transition, returns VRAM requirement, speedup estimate, and approximate quality delta.

  Parameters:

  • params_billions (required) — number — Model parameter count in billions
  • batch_size — number — Serving batch size (default 1)
  • precision_to — string — Target precision (default int4)
  • precision_from — string — Starting precision (default bf16)
  • kv_cache_tokens — number — Max KV cache tokens (default 8192)

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• fine-tune-roi-calculator read-only

  Use when a user is considering fine-tuning vs prompt engineering. Returns training cost, monthly inference savings, months-to-ROI, and breakeven volume.

  Parameters:

  • train_tokens (required) — number — Training tokens (dataset × epochs)
  • monthly_inference_tokens (required) — number — Expected monthly inference volume (output tokens)
  • train_cost_per_1m — number — Training cost per 1M tokens
  • prompt_reduction_pct — number — Prompt size reduction % from eliminating few-shot (default 0)
  • base_inference_cost_1m — number — Baseline API output cost per 1M
  • finetuned_inference_cost_1m — number — Fine-tuned inference cost per 1M

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• eval-cost-calculator read-only

  Use when a user needs to budget an LLM evaluation run. Given samples/models/trials, returns total cost, per-run cost, and parallel time estimate.

  Parameters:

  • samples (required) — number — Number of eval samples
  • model — string — Eval model (default claude-sonnet-4-6)
  • models — number — Candidate models (default 1)
  • provider — string — Eval provider (default anthropic)
  • avg_tokens_in — number — Avg input tokens per sample (default 2000)
  • judge_enabled — boolean — Enable LLM-as-judge second pass (default false)
  • avg_tokens_out — number — Avg output tokens per sample (default 500)
  • judge_tokens_in — number — Judge input tokens (default 1500)
  • judge_tokens_out — number — Judge output tokens (default 200)
  • trials_per_sample — number — Repeats per sample (default 1)

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• observability-cost-calculator read-only

  Use when a user needs to budget LLM observability tooling. Returns monthly cost at given request volume with retention adjustment.

  Parameters:

  • requests_per_day (required) — number — Average daily LLM requests
  • provider (required) — string — Observability provider
  • avg_log_bytes — number — Avg payload bytes per traced request (default 4096)
  • retention_days — number — Retention in days (default 30)

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• rag-pipeline-cost-calculator read-only

  Use when a user needs end-to-end RAG cost estimation (embedding + vector store + generation). Returns monthly cost with breakdown and dominant-component identification.

  Parameters:

  • queries_per_day (required) — number — User query volume per day
  • corpus_tokens (required) — number — Indexed corpus size in tokens
  • vector_store — string — Vector store
  • answer_tokens — number — Avg answer tokens (default 400)
  • generator_model — string — Generator model (default claude-sonnet-4-6)
  • question_tokens — number — Avg question tokens (default 100)
  • chunks_retrieved — number — Chunks per query (default 5)
  • chunk_size_tokens — number — Avg chunk size (default 512)
  • embedding_provider — string — Embedding model
  • generator_provider — string — Generator LLM provider (default anthropic)
  • reindex_fraction_per_month — number — Fraction of corpus re-embedded per month (default 0.1)

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• automation-cost-calculator read-only

  Use for generic or non-agent recurring workflow platform billing across Zapier task billing, Make credit billing, and n8n execution billing. Use agent-workflow-cost-calculator instead when the workflow is explicitly an AI agent with app actions or MCP tool calls. Use agent-loop-cost-calculator instead when the question is about LLM inference/reasoning cost per successful agent task.

  Parameters:

  • workflow_runs_per_month (required) — number — Workflow/scenario runs per month
  • make_modules_per_run — number — Make module actions per scenario run; defaults to billable_steps_per_run
  • billable_steps_per_run — number — Billable actions/modules per run (default 5)

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• agent-workflow-cost-calculator read-only

  Use for automation-platform/iPaaS cost of an AI agent workflow, including app-action fan-out and MCP tool-call accounting, explicitly excluding LLM token spend. Use automation-cost-calculator for generic non-agent workflow billing. Use agent-loop-cost-calculator for LLM inference/reasoning cost per successful multi-step agent task.

  Parameters:

  • agent_runs_per_month (required) — number — Agent workflow runs per month
  • app_actions_per_run — number — Downstream app actions per agent run (default 3)
  • make_modules_per_run — number — Make modules per run; defaults to app actions + MCP calls
  • mcp_tool_calls_per_run — number — MCP tool calls per agent run (default 1)

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• agent-loop-cost-calculator read-only

  Use for LLM inference/reasoning cost per successful multi-step agent task, including failure overhead and context growth across turns. Use agent-workflow-cost-calculator for iPaaS/app-action/MCP orchestration fees. Use automation-cost-calculator for generic Zapier/Make/n8n workflow billing.

  Parameters:

  • tasks_per_month (required) — number — Tasks attempted per month
  • model — string — LLM model (default claude-sonnet-4-6)
  • provider — string — LLM provider (default anthropic)
  • success_rate — number — Task success rate 0-1 (default 0.7)
  • steps_per_task — number — Avg reasoning steps per task (default 5)
  • tool_calls_per_step — number — Avg tool invocations per step (default 2)
  • context_growth_factor — number — Multiplier on input tokens as conversation grows (default 1.4)
  • avg_tokens_in_per_turn — number — Avg input tokens per turn (default 3000)
  • avg_tokens_out_per_turn — number — Avg output tokens per turn (default 400)

  Hints: read-only yes · destructive no · idempotent yes · open-world no

Publisher website

Measured by MCPYP probing. Untested fields are omitted, never estimated. Listing is always free; MCPYP never proxies calls or stores user tokens.